Brain responses are dynamic, state dependent, and change over time. In acknowledging this, it is becoming increasingly more important to characterize single trial responses. Using Bayesian probability theory, I demonstrate how a more complex signal model that accounts for amplitude and latency variability in the single trial allows one to characterize and examine single trial responses. Furthermore, the fact that different neural ensembles exhibit differential variability patterns allows one to tease apart these mixed signals and isolate signals generated from different neural ensembles. This new technique, called differentially Variable Component Analysis (dVCA) allows one to separate, identify, and characterize single trial responses from simultaneously active sources. Interactions among the ensembles can be studied by examining their single trial properties. I will demonstrate the technique using simulations as well as intracortical field potentials recorded from a linear multi-electrode array in a macaque experiencing visual stimulation.